Eric Horvitz

Adjunct Professor@Columbia University·美国·OpenReview

研究方向

human-AI interaction and collaboration · augmenting human intelligence · decision-theoretic inference · utility-theoretic models · Markov decision processes · MDP · POMDP · cognitive science · human cognition · cognitive architectures · bounded rationality · bounded optimality · self-supervised training · semi-supervised training · causal inference · AI safety · reliability · robustness · adversarial machine learning · reward hacking · Bias · Fairness · Explanation of machine learning · transparency · Lifelong machine learning

4.7

Tracing the Traces: Latent Temporal Signals for Efficient and Accurate Reasoning

ICLR 2026Poster

4.0

Utility-Directed Conformal Prediction: A Decision-Aware Framework for Actionable Uncertainty Quantification

ICLR 2025Poster

4.3

MedFuzz: Exploring the Robustness of Large Language Models in Medical Question Answering

合作者 (20)

Vidhisha Balachandran

Eric Horvitz

Tracing the Traces: Latent Temporal Signals for Efficient and Accurate Reasoning

The Collaboration Gap

Magentic Marketplace: An Open-Source Environment for Studying Agentic Markets

Improving Instruction-Following in Language Models through Activation Steering

Utility-Directed Conformal Prediction: A Decision-Aware Framework for Actionable Uncertainty Quantification

MedFuzz: Exploring the Robustness of Large Language Models in Medical Question Answering